Viewpoint on Multisensory Perception for Action

Abstract In this viewpoint, Laurence Harris and Marc Ernst reflect on their careers studying how the brain combines information from multiple sensory systems — vision, hearing, touch, and the vestibular sense — to perceive the world and guide action. They trace the field’s evolution from early psychophysical and neurophysiological work to the introduction of Bayesian models, which reframed multisensory integration as a problem of statistical inference under uncertainty and provided a principled link between behavioural findings and normative theory. Both authors highlight open challenges: bridging computational, algorithmic, and neural levels of explanation; understanding how the brain represents and approximates complex, context-dependent probabilistic information; and characterising the naturalistic sensory signals generated during active behaviours such as walking, grasping, and navigating. They discuss the growing role of virtual reality in enabling more ecologically valid experiments while preserving experimental control and note that progress has been uneven across modalities — most advanced in vision, less so in audition and haptics. The authors also address open-science practices, the value and limits of replication, and the anticipated role of artificial intelligence as a tool that accelerates research workflows rather than replaces scientific reasoning. They close with reflections on mentorship and curiosity as drivers of scientific careers, and a shared conviction that no sense operates in isolation — multisensory integration is fundamental to perception and action and remains one of the most promising frontiers in cognitive science.